question about 13900k degradation

[tomerturbo]

hello everyone i hear a lot people say the 13900k degradation

so i want to know what the main reason that cause degradation

i have asus z690 hero and 13900k my setings is P-CORE 56 E-CORE 45 RING 50 auto voltage llc3 DC_LL 1.10 AC_LL 0.20 my load voltage in r23 1.18v cpu power 260w my

cpu package temp is 92c does it ok for cpu run this setings can it cause degradation fast?

thanks for the help

 

[ir_cow]

If you read TPU CPU reviews, Intel says 100c is perfectly fine (12 and 13th gen). For the KS it's rated to 110C.

Degradation can came from temps if it's outside the operating specifications, but usually the voltage is what does the CPU in.

 

[GerKNG]

degrading the actual cores is extremely hard.
if you want to lose 20Mhz you probably need to run Prime95 small fft with AVX2 at 115°C and 1.45V+ Load VCore with a flat LLC for several days.
run (stock without PL) R23 and look what the VCore is.
this +5% is fine for basically ever.
your ~1.2V settings will probably outlast you before the chip would degrade from that.

What you should care about is the IMC/System Agent Voltage.
Memory Controllers degrade pretty fast with high voltages. (VCCSA should be below 1.4V for long term. (mine is at 1.2V stock with 32GB 6000 CL30)

 

[Bloax]

What you should care about is the IMC/System Agent Voltage.
Memory Controllers degrade pretty fast with high voltages. (VCCSA should be below 1.4V for long term. (mine is at 1.2V stock with 32GB 6000 CL30)

The fastest way to degrade the memeory controller besides slamming it with 1.5v and plenty of activity, is actually not using an LGA1700 Contact Frame replacement, as the CPU will "by default" get bent in such a way by the stock CPU-fastening mechanism that it loses contact with the memory interface pins.

You can read a little adventure on this from end-to-beginning (conclusions are a nice tl;dr) right about here: https://www.overclock.net/threads/m...i-owners-thread.1795717/page-47#post-29030497



The ring frequency might degrade if you run it at its maximum while slamming 1.4v into the CPU, just as it did with Comet Lake (skylake+++, the 10900k & frens generation), it might not.
Besides that you shouldn't really worry about it, and if you want the CPU to not get hot - simply disable hyperthreading.

 

[phanbuey]

I haven't heard anyone talk about degradation for a while now tbh. Since about 7th gen i've always run my intel chips into the 90's and pretty aggressive OC's with high SA volts ~1.38v / 4 sticks of ram - and they've been stable no real degradation.

I've had a ram stick die on me, but that's about it.

 

[ir_cow]

The fastest way to degrade the memeory controller besides slamming it with 1.5v and plenty of activity, is actually not using an LGA1700 Contact Frame replacement, as the CPU will "by default" get bent in such a way by the stock CPU-fastening mechanism that it loses contact with the memory interface pins.

While I agree the ILM sucks that can cause memory problems with certain coolers and motherboards, I'm not sure about your contact frame claims that without one your CPU degrades.

 

[RJARRRPCGP]

If you read TPU CPU reviews, Intel says 100c is perfectly fine (12 and 13th gen). For the KS it's rated to 110C.

Degradation can came from temps if it's outside the operating specifications, but usually the voltage is what does the CPU in.

Even more so, a combination of a lot of Vcore overvolting and high core temps! That also seems to cause the phenomenon of being unstable at higher than 55C, at least in some cases!
And very well can be unstable at only 65C! Symptoms to look for: 1: hard lock crash with a heavy load or an error with a heavy load. Most often an error from the stress tester or a WHEA BSOD.
Possibly "IRQL_NOT_LESS_OR_EQUAL" / A.K.A. "0x0000000a" BSOD as well.

 

[ir_cow]

Even more so, a combination of a lot of Vcore overvolting and high core temps! That also seems to cause the phenomenon of being unstable at higher than 55C, at least in some cases!

Thats just as bad CPU. RMA it.

 

[RJARRRPCGP]

Thats just as bad CPU. RMA it.

The symptoms go away when returned to stock or less of an OC, when I had '00s CPUs do that. (especially socket 462!)

 

[ir_cow]

The symptoms go away when returned to stock or less of an OC, when I had '00s CPUs do that. (especially socket 462!)

First off you didn't say anything about a OC, and second, your OC is just unstable.

 

[fevgatos]

The symptoms go away when returned to stock or less of an OC, when I had '00s CPUs do that. (especially socket 462!)

Lower temps require less voltage for stability. Thats why when you are overclocking you should run some stability tests with fans turned off to hit 90c. At least thats my method

 

[Bloax]

While I agree the ILM sucks that can cause memory problems with certain coolers and motherboards, I'm not sure about your contact frame claims that without one your CPU degrades.

I don't need to "be sure" when my maximum DDR4 memory frequency went from ~4100 (""mt/s"") when I first assembled the system, down to around 4040 eight months later, then back to ~4100 after enough trial & error with tiny turns of a thermal grizzly plate. (With plenty of fails after nailing it first try, just for good measure)

Then other people experienced similar issues, and had a similar solution to this issue.

If I can't "be sure" having lived it, and having other people live it too, then I would not be sure of anything at all - and that is no way to live!

 

[ir_cow]

I don't need to "be sure" when my maximum DDR4 memory frequency went from ~4100 (""mt/s"") when I first assembled the system, down to around 4040 eight months later, then back to ~4100 after enough trial & error with tiny turns of a thermal grizzly plate. (With plenty of fails after nailing it first try, just for good measure)

way to many factors here you aren't considering to make a absolute claim like you are doing.. What was the CPU SA,VDD2 and TX voltages before? Is the sub-timings the same. How about Gear 1 vs Gear 2. Was it even stable in the first place?

A Contact frame is good to create better pin contact. A few outlets do have a good detailing why the ILM frame can be a issue for higher memory freqs, with the bonus of lowering the CPU temp a few degrees. Still not really "degradation". degradation is when it becomes permanent. Like needing more voltage for the same results.

 

[Bloax]

Yes, yes, armchair expertise in throwing around entry-level terms and babby's first dum-dum as creme-de-la-creme examples in "failure to control variables". (VDD2 isn't on ddr4, it's for ddr5 - by the way)
I am very convinced that people who have spent entirely too much time and brain cells grokking memeory failed to control any variables, in spite of this hobby demanding you Do That Or Fail! Laughing emoji!

In fact due to book-keeping, I can tell that somewhere between June 2022 and August 2022 the maximum 2x8 frequency went from 4194 to 4166, to around 4140 - with a screenshot of 4126 being run to avoid "near future" instability.
I didn't notice the specifics, as I was busy running a mixed 4x8 config while being IMC-capped rather than frequency-capped.

Usually I only test for a couple of hours, as anything beyond that tends to run into Random Power Grid Voltage Drop Under Full RAM Load errors.

I tend to run my memory rather stable, but somehow I am the one the doubt is being cast on.
I don't doubt that my "fellow memory OC enjoyers" are a little crazy, and that when they claim to measure their CPU for bent-ness, they actually did so - and consequently have a point to it being bent after long-term LGA1700 mounting versus a fresh CPU.

there's this unspoken rule amogus of always entering all relevant timings so that the board doesn't muck-up things by chance, so as to keep things consistent between attempts to Run Something
or, as it's otherwise called, there's this unspoken rule of controlling variables religiously hee hee

I do, however, doubt when somebody casts doubt over obvious consequences of failing memory interface pin contact due to (cost-saving) design flaws causing bending in that area.
And that solving this issue, so as to not run into the consequences of bad memory interface pin contact; reportedly, potentially as severe as ddr4-3600 XMP auto settings eventually not working
- supposedly Old Wives' Tales, total fabrication by The Official AMD Fanboys' Club (of which I am a proud member of, as well as the Intel Enjoyers' Club, triple agent!) to spook people from buying those processors.



but really i am just wasting my breath here and you're lucky i was heading to bed and do stupid things like this waste of effort on my way back to good nighty-nights land

 

[ir_cow]

@Bloax I'm not saying your claim that you can no longer run what you did at first is false. What I am saying is, while you may think the contact frame is the sole reason that a higher memory works again is false. It may play a role because it often provides better pin contact. What you claiming is probably due to a slight degradation and (maybe) the contact frame now just provides better contact. Without knowing the details, I assume you left all your CPU voltages on auto, Its more likely that contact frame actually has little to do with what your experience and it just coincides instead. A false positive if you will.

 

[Deleted member 185158]

@Bloax I'm not saying your claim that you can no longer run what you did at first is false. What I am saying is, while you may think the contact frame is the sole reason that a higher memory works again is false. It may play a role because it often provides better pin contact. What you claiming is probably due to a slight degradation and (maybe) the contact frame now just provides better contact. Without knowing the details, I assume you left all your CPU voltages on auto, Its more likely that contact frame actually has little to do with what your experience and it just coincides instead. A false positive if you will.

Bro. Memory degradation is not the same as cpu. The contact frame doesn't magically make memory overclocking better. I don't run a contact frame. The pins contact the cpu fine. 13600KF, 13700K both of which my memory acts just the same with either chip.

99% of the non extreme cooling users should never have cpu degradation concerns.

 

[ir_cow]

@ShrimpBrime I'm not saying memory degradation is the same as CPU degradation.... Also a contact frame doesn't magically make things better either

 

[freeagent]

Thermalright sent me a blue and black contact frame.. I guess I should use one in a new system sometime

Edit:

Meaning this sounds like fun..

 

[Deleted member 185158]

@ShrimpBrime I'm not saying memory degradation is the same as CPU degradation.... Also a contact frame doesn't magically make things better either

That was an agreement statement. Bloax is arguing memory capability because of a contact frame; however the frame is to prevent board flex and pins not touching. Either the pins are touching or they are not. This has nothing to do with CPU degradation. I suppose you could degrade a memory controller, in which case you may loose dual or quad channel, or would have a need for increased IMC voltage for a certain clock speed after a given time of proven stability.... But memory (in my opinion) isn't always stable after a restart if and when you're tweaking well beyond specifications.

There's really no "GOOD" answer for CPU degradation. I have CPUs (as others do) that are 15-20 years old. Where used once daily. Myself having used extreme cooling on several processors as old as Socket A and 754. And ironically, I still have some of this hardware and try to use them once a year. Not for the processors, but for the motherboards. Gotta charge those old caps once in a while ya know??

But it seems, every processor release there's questions regarding degradation.

On average, most electronics (including processors) are rated with a MTBF of 100,000 hours at full load use. That's 10 years roughly.

Degradation worries? Nah.
Can you break the hardware? Yes indeed.
Does degradation happen faster if I overclock beyond specifications? Yes.
Do you need to overclock your gaming rig's processor in 2023? No.
Do you need to overclock your Barton core in 1999? Yes. (lol)
If I turn it on and use it at defaults, can I expect 10 years from my CPU?? YES.

 

[fevgatos]

Yes, yes, armchair expertise in throwing around entry-level terms and babby's first dum-dum as creme-de-la-creme examples in "failure to control variables". (VDD2 isn't on ddr4, it's for ddr5 - by the way)
I am very convinced that people who have spent entirely too much time and brain cells grokking memeory failed to control any variables, in spite of this hobby demanding you Do That Or Fail! Laughing emoji!

View attachment 282229

In fact due to book-keeping, I can tell that somewhere between June 2022 and August 2022 the maximum 2x8 frequency went from 4194 to 4166, to around 4140 - with a screenshot of 4126 being run to avoid "near future" instability.
I didn't notice the specifics, as I was busy running a mixed 4x8 config while being IMC-capped rather than frequency-capped.

Usually I only test for a couple of hours, as anything beyond that tends to run into Random Power Grid Voltage Drop Under Full RAM Load errors.
View attachment 282230
View attachment 282231
I tend to run my memory rather stable, but somehow I am the one the doubt is being cast on.
I don't doubt that my "fellow memory OC enjoyers" are a little crazy, and that when they claim to measure their CPU for bent-ness, they actually did so - and consequently have a point to it being bent after long-term LGA1700 mounting versus a fresh CPU.

there's this unspoken rule amogus of always entering all relevant timings so that the board doesn't muck-up things by chance, so as to keep things consistent between attempts to Run Something
or, as it's otherwise called, there's this unspoken rule of controlling variables religiously hee hee

I do, however, doubt when somebody casts doubt over obvious consequences of failing memory interface pin contact due to (cost-saving) design flaws causing bending in that area.
And that solving this issue, so as to not run into the consequences of bad memory interface pin contact; reportedly, potentially as severe as ddr4-3600 XMP auto settings eventually not working
- supposedly Old Wives' Tales, total fabrication by The Official AMD Fanboys' Club (of which I am a proud member of, as well as the Intel Enjoyers' Club, triple agent!) to spook people from buying those processors.



but really i am just wasting my breath here and you're lucky i was heading to bed and do stupid things like this waste of effort on my way back to good nighty-nights land

Just a personal anecdote, my unify X was running 6000c30 with tight subs for almost a year. Then one day, it failed to boot repeatedly, no matter what. I thought it was indeed due to bending. It wasn't,, I updated the bios and voila, it instabooted with the same timings I was running before, probably bios corruption of some kind. As ir_cow said, the variables are SO many. Not saying it definitely isn't due to bending, im saying it's really hard to show that it is due to bending